OpenAI-compatible - Bed of Roses

The OpenAI-compatible provider connects BOR to any endpoint that speaks the OpenAI Chat Completions API: hosted vendors, gateways, or local model servers.

Quick presets

The setup panel has one-click preset chips that fill in the base URL and a sensible model:

Preset	Base URL	Example model
DeepSeek	`https://api.deepseek.com`	`deepseek-v4-pro`
Kimi (Moonshot)	`https://api.moonshot.ai/v1`	`kimi-k3`
Z.ai (GLM)	`https://api.z.ai/api/paas/v4`	`glm-5.2`
Groq	Groq endpoint	`llama-3.3-70b-versatile`
OpenRouter	OpenRouter endpoint	`anthropic/claude-sonnet-5`
Together / Mistral / Perplexity / Fireworks / DeepInfra / Cerebras	vendor endpoints	vendor models
Ollama (local)	`http://localhost:11434/v1`	no key needed
LM Studio (local)	`http://localhost:1234/v1`	no key needed

Setup

In onboarding (or provider Settings), pick OpenAI-compatible.
Click a preset — or enter a base URL by hand.
Enter an API key if the endpoint needs one (local servers usually don’t).
Enter the model name exactly as the endpoint serves it, and pick a reasoning effort.

Notes

There’s no fixed catalog — the model name is whatever your endpoint serves.
Vision and image support depend on the underlying model; text-only models can’t use screenshot-driven features.
Reasoning effort maps through where the endpoint supports it (e.g. DeepSeek’s reasoning models).

Troubleshooting

Connection refused — the server isn’t running or the base URL/port is wrong.
404 / unknown model — the model name doesn’t match what the endpoint serves.
No streaming — the endpoint must support OpenAI-style streamed responses.
See Troubleshooting.